6 replies to this topic

[Elizander]

If PSR is determined by match score and PSR is used to match players, then it'd be reasonable to think that match score and all its related bonuses is the core of the matchmaker system.

While match score works fine for rewards, does it really work well enough to determine a players Tier? This isn't about me or anyone else wanting to be in T1. I've stated many times that I'm more than fine being in T3/T4 and reaching T3 if I drop to T4 is very easy for me. This is about getting a better understanding of what rewards or data from players should be used to determine if they did well in a match or not.

I understand that there are many factors that goes on in a match and it's not something that you can always quantify in numbers. The scoring system also can't be some complex data science AI that runs machine learning algorithms on full match data after each match is played out before giving out rewards. I suppose we can at least try to put together what actions and the timing of those actions players take which should at least be considered for match score.

In general, what actions lead to a win? Dealing damage, sure. Killing an enemy mech, check. Sticking with your team? Ok. That's the general gist and measure of performance in the game right now.

• The goal is to win.
  
• Killing enemy mechs helps you win.
  
• Dealing damage causes enemy mechs to die.
  
• Sticking together helps your team deal more damage to enemy mechs with focus fire.
  
• Shooting down missiles with AMS reduces the chance of the other team killing your mechs.

For the most part I think that's the basic pillar of performance or skill determination in this game. So what actions do you think should be rewarded or considered for match score in order to improve matchmaking?

I'm thinking about things like:

• Killing enemy mechs before they can deal significant damage - Shutdown
  
• Dealing efficient damage for quick kills to CT/Head - Core Damage / Headshot Kill
  
• Shooting back while under fire - Trading Blows
  
• Initiating a push while with other teammates - Spearhead

Those are just quick ideas off the top of my head. I'm no 1%er on Jarls. If you think there are other actions that should be considered for match score and maybe other actions that should be rewarded less then you can post it here.

[Aidan Crenshaw]

Matchscore already takes most of the things you mentioned into account, except for

Elizander, on 20 January 2021 - 04:40 AM, said:

• Initiating a push while with other teammates - Spearhead

and somewhat

Elizander, on 20 January 2021 - 04:40 AM, said:

• Dealing efficient damage for quick kills to CT/Head - Core Damage / Headshot Kill

While you are not rewarded directly for clean kills, indirectly you'll get solo kill.
The factors that PGI couldn't reward directly are now calculated into PSR change. It's their idea that all the little things like tanking for the team, making calls and so forth result in a more likely win. That's why a winner's matchscore is weighted a little higher than a loser's matchscore when it comes to PSR change.

[VonBruinwald]

Aidan Crenshaw, on 20 January 2021 - 06:30 AM, said:

The factors that PGI couldn't reward directly are now calculated into PSR change. It's their idea that all the little things like tanking for the team, making calls and so forth result in a more likely win. That's why a winner's matchscore is weighted a little higher than a loser's matchscore when it comes to PSR change.

But if that's the logic why not drop MS entirely and base it on Win/Loss? Win/Loss is the ultimate measure of how well your efforts contribute to the teams performance.

As it stands, a squirrel who draws 5 enemy mechs from the front and dies with double digit damage is a "worse" performer than the light who avoids the battlefield to farm an AFK for damage.

[Haipyng]

VonBruinwald, on 20 January 2021 - 08:39 AM, said:

But if that's the logic why not drop MS entirely and base it on Win/Loss? Win/Loss is the ultimate measure of how well your efforts contribute to the teams performance.

As it stands, a squirrel who draws 5 enemy mechs from the front and dies with double digit damage is a "worse" performer than the light who avoids the battlefield to farm an AFK for damage.

That goes along with the rewards scheme in MWO. Achieving objectives is so far under rewarded as to be non-existent. This is why every map is played as skirmish (and people actually get pissed at you for taking advantage of the win objectives).

A player's worth or skill is not measured purely in the PSR. Let's not forget there are safeties built into MM that let higher tier players in with lower tier players anyway, in the name of speedy matches.

Of course there are also tonnage considerations (more weight is more health and potentially more DPS). Mechanics (a build that plays favorably on a map, like long/short range, hot or cold maps, terrain) also plays a role. MM does not consider these things at all. I have been on plenty of maps with one team with 6 assaults and the other team with one, or one team all long range dakka on Terra and the other short-med range MPL and SRMs, or lots of LRMs and ECM on one side on Polar.

PSR is a very small unreliable indicator of a player's value that could be rendered moot by the safety and certainly does nothing to balance the rest. The tonnage check was removed years ago. There has never been a check on mechanics.

Given the small player base we have today, it is impossible to have MM balance on any meaningful criteria at all.

Edited by Haipyng, 20 January 2021 - 09:42 AM.

[Vlad Ward]

The one thing about MS I might change would be weighting match score earnings from damage based on the component damage is applied to. Could maybe make that a little more interesting by also adding another mod which checks to see if the component holds equipment, to reward ripping off weapon arms. PPC boats who actually land hits on CTs (which carry engines) could be worth more than LRM boats that scrape armor off shield arms.

Edited by Vlad Ward, 20 January 2021 - 10:54 AM.

[GARION26]

VonBruinwald, on 20 January 2021 - 08:39 AM, said:

But if that's the logic why not drop MS entirely and base it on Win/Loss? Win/Loss is the ultimate measure of how well your efforts contribute to the teams performance.

As it stands, a squirrel who draws 5 enemy mechs from the front and dies with double digit damage is a "worse" performer than the light who avoids the battlefield to farm an AFK for damage.

While there are some who favor a WLD ratio driven mechanic I think a simple WLD model is doomed for failure (a highly complicated one might work.)

If the WLD matchmaker works - it causes regression to the mean on your primary matchmaking statistic. You win 0.6 rate for 10 matches you will get better opponents - and probably drop down closer to 0.5 total ratio over your next 20 games. Someone else has a 0.4 over 10 games they get easier matches and get close to 0.5. Those two players would be in a match, make those 50 or 100 or 1000 match records and again you will continually be pushing some very skilled players primary matchmaking characteristic (WLD) down towards the 0.5 mean, and some very bad players up towards 0.5.

The primary criteria a simple WLD system would use won't differentiate how folks got to their current WLD and regresses people to look like they are equally skilled even when they aren't.

An incremental system (which PSR is) is better at capturing your longer term track record. PSR and match score aren't perfect and aren't idiot proof (you can engage in 'economically perverse behaviors' that artificially raise your PSR while worsening your WLD if you really want to) but they aren't terrible approximations of the pretty complicated factors that result in higher win ratios.

As always I say just check Jarl's out and sort by WLD and look at the data we have on highly successful pilots - the kind of things they tend to do well in match (KD, damage) track very well with who does well on WLD and who tends to do well on Matchscore by the currrent system. You have to ignore the small sample size folks of course (less then 20 games for example.)

https://leaderboard....g/?o=wlr&d=DESC

Edited by GARION26, 21 January 2021 - 09:35 AM.

[Dangit Karen]

Vlad Ward, on 20 January 2021 - 10:52 AM, said:

The one thing about MS I might change would be weighting match score earnings from damage based on the component damage is applied to. Could maybe make that a little more interesting by also adding another mod which checks to see if the component holds equipment, to reward ripping off weapon arms. PPC boats who actually land hits on CTs (which carry engines) could be worth more than LRM boats that scrape armor off shield arms.

You mean reward meaningful damage? >gasp<

Even though I am a potato, I 100% AGREE with you.